Participants: Bryce Tyner (NCSU), Michael Brennan (NHC), Reid Hawins (ILM), Jonathan Blaes (RAH), David Glenn (MHX), Frank Alsheimer (CHS)
The CSTAR TC Winds Team held their monthly conference call on Wednesday, 3/13/13. Bryce began the meeting by noting that the Holland et al. verification images have been posted to: http://www4.ncsu.edu/~bptyner/holland.html. Collaborators are encouraged to look at the images and provide feedback to Bryce on any observed interesting features from the plots.
Bryce summarized the methodology used in the hybrid approach to an improved interpolation scheme. The method uses the error minimization technique. Error was calculated as the difference between the H*Wind analyses and the Holland et al, with positive values indicating positive bias and negative values indicating negative bias. The error was normalized relative to the NHC best track maximum sustained winds for each analysis time in order to allow for inter-storm comparisons. Normalized error was plotted as a function of radius for each quadrant and each analysis time. Some of the key results include:
-For all four quadrants, a large negative difference between the Holland et al. and H*Wind wind fields was observed right near the center of the storm. This was likely a product of the Holland et al. interpolation method, which assumes a near linear increase in wind speeds from the center of the storm to the radius of maximum winds.
-Near the radius of maximum winds, there was an observed peak positive difference between the Holland et al. and H*Wind wind fields, suggesting a positive bias in the interpolated wind fields. This is consistent with the Holland et al. (2010) analysis, where the authors indicate that the positive bias is the product of sharp wind gradients near this radius of maximum winds. It is also likely partially attributable to sampling issues for the H*Wind analyses, which may not have surface observations at the locations of peak winds. Finally, the Holland et al. (2010) model assumes peak winds to be observed in all four quadrants and radial uniformity within each quadrant at each distance form storm center. This is likely an unrealistic assumption when compared to real tropical cyclones.
-Beyond radius of maximum winds, the error tends to decay to a value near 0, then eventually becomes negative at around 200 km from storm center. This negative value rebounds to slightly positive error after a distance of approx. 300 km from storm center. There is large clustering in the errors as a function of radius as seen in the plots.
It is encouraging that there indeed appears to be a systematic relationship between error and distance from storm center. An error function is currently being developed using least squares fit. The empirically-based function will then be added to the raw Holland et al. interpolated wind field and a comparison to previously interpolated wind speeds will be developed.
Jonathan concluded the call by discussing the main points of the recent National Hurricane Team conference call. The next National Hurricane Team conference call will include a discussion of the CSTAR TC Winds research and the proposed testing of the TCMWindGustTool.
The next CSTAR TC Winds conference call will be held on Wednesday, April 10th at 11 AM.